Technology
When software gets faster, judgement matters more
A feature that once needed several days can now arrive as a working prototype before the meeting ends. The speed is useful. It also makes it easier to avoid the harder question of whether the feature should exist.
By FABCODE2 min read

Implementation is only one part of delivery
A business requirement usually contains decisions that no code-generation tool can settle on its own. Who is allowed to change a record? Which system owns the customer relationship? What happens to an order when two services disagree? These questions determine whether software fits the operation it is supposed to serve.
DORA's 2025 research describes AI as amplifying an organisation's existing strengths and weaknesses, and points to the surrounding organisational system as central to realising its value. The implication for a company builder is straightforward: introducing faster implementation into an unclear process will not resolve the unclear process.
An illustrative customer dashboard makes the point. Generating its interface may be relatively quick. Deciding what the revenue figure includes, when a customer becomes active and which exceptions an account manager should see is a separate piece of work. If those definitions are wrong, a polished dashboard merely presents disagreement more efficiently.
The scarce resource moves towards decisions
Lower implementation costs should make teams more selective, because they can test an idea earlier. Yet they can also encourage a larger queue of almost-finished features. Each addition creates something to secure, support, explain and eventually replace. The maintenance obligation survives the excitement of the first demonstration.
A sensible response is to keep experiments small and decisions explicit. Before building, write down whose behaviour should change and what evidence would justify continuing. After the experiment, decide whether to deepen it, revise it or remove it. Cheap implementation is especially valuable when it makes an unpromising idea cheap to abandon.
Verification deserves the same attention. Tests can establish that a calculation matches an agreed rule; they cannot establish that the rule is commercially sensible. Review should distinguish between implementation correctness and whether the software solves the intended problem. Both need an accountable owner.
Read the delivery system, not the demonstration
When assessing a technology business, a rapid demonstration is a starting point. More revealing questions concern the next release. Can another engineer understand the change? Can the team deploy and reverse it safely? Does customer feedback reach the people making product decisions? Are support issues narrowing, or merely accumulating behind new features?
There is no virtue in preserving slow work simply because it used to require expertise. Teams should use tools that remove unnecessary effort. They should also recognise where effort has moved: into defining the problem, choosing boundaries and checking consequences.
The opportunity is to turn shorter implementation cycles into shorter learning cycles. A company that makes better decisions each week can compound that advantage. A company that merely produces more software may be increasing its future workload.


